Jaggies in D-400Z photos

greenspun.com : LUSENET : Imaging Resource Discussion : One Thread

Has anyone ever noticed "jaggies" (staircase effect on edges) in photos from the Olympus D-400Z? It seems that on occasion, images have what I think is an excessive amount of jaggies. Here is one photo which shows the problem: http://albums.photopoint.com/j/View?u=27007&a=153394&p=15764882&Sequence=0 This photo was taken in SQH mode. (1280x960, using the lowest amount of compression possible, before going to uncompressed TIFF). It has been rotated, and cropped (my girlfriend doesn't like her photo sprayed around the net) on the left hand side, in Photoshop 5, and saved at JPEG quality level 8. I assure you that the jaggies look identical in the original photo.

If you look at the test shots right here at imaging-resource, the same effect is present, although not to the same extent as the above shot. Even the D-450Z exhibits the jaggies. (look at the first outdoor portrait photo taken with the D-450Z - jaggies are clearly visible).

I'm trying to create a "test scene" at home, to see whether the problem goes away when using uncompressed TIFF. So far I haven't been successful in creating a suitable test scene, though. (whenever I *want* the jaggies to occur, they don't seem to!)

Anyone else notice this? Does it go away when using uncompressed TIFF?

Greg.

-- Greg Sullivan (gregory.sullivan@compaq.com), January 30, 2000

Answers

Don't know about the rotated vs unrotated issue, but FWIW, I've noticed increased jaggies on many cameras along blue/red boundaries. This is doubtless because the most common "striping" pattern for applying the color filters to the CCD surface uses twice as many green as red or blue filters. (They do this because the green channel tends to follow the "luminance" channel in our eyes more closely, so they get better apparent resolution.) The result is that there are fewer red/blue pixels on the sensor, so when you encounter a boundary between those colors, you're much more likely to see jaggies. In our outdoor portrait shot, you'll probably see jaggies on a lot of cameras between the red and blue flowers, although there are clear differences from model to model.

-- Dave Etchells (web@imaging-resource.com), January 30, 2000.

Oops - here is the correct URL for my D-400Z photo which shows the jaggies: http://albums.photopoint.com/j/View? u=27007&a=459989&p=15764882&Sequence=0

-- Greg Sullivan (gregory.sullivan@compaq.com), January 30, 2000.

Interesting. The original, *unrotated* image, exhibits less jaggies than the rotated version. There are still jaggies present, but in my opinion they are less severe. I've posted the unrotated image at:http://albums.photopoint.com/j/View? u=27007&a=459989&p=15770086&Sequence=0 Note that it has been cropped in the same way as the first example.

The URL for the "album" is: http://albums.photopoint.com/j/AlbumIndex?u=27007&a=459989

I can't explain the difference. A 90 degrees rotation should look identical, because the pixels on our monitors are basically square. Is it perhaps something to do with the fact that the camera somehow optimises the image so that it looks best when viewed horizontally, at least, on standard (horizontally scanning) computer monitors? Note that I've tried two different programs to do the rotation - Photoshop 5.0 and Microsoft PictureIT 99. In Photoshop, if I upsample a lot first (4X, for example), rotate it, then downsample to the original resolution, the result looks better than a simple rotation. This might just be because the upsampling and downsampling is for some reason smoothing out the jaggies - it may not represent a *true* increase in quality. There's no question that it looks *better* though.

Greg.

-- Greg Sullivan (gregory.sullivan@compaq.com), January 30, 2000.


Thanks Dave. Yes, it would make perfect sense for there to be more jaggies on red/blue boundaries. However, I have looked at a lot of the sample outdoor portrait photos from circa 1Megapixel cameras, and it seems to me that most of them do not exhibit this problem. (to my untrained eyes, anyway). In fact, I can only really see jaggies in the D-400Z, the D-450Z, and the Coolpix 900. The Coolpix 900 jaggies were bad enough to draw comment in the review of the photo, in fact. (it does appear to be the "worst", although I appreciate that I am only looking at ONE aspect of the photos at the moment. Overall the Coolpix 900 does of course take fantastic photos!)

I've found that if I downsample to approximately half the total number of pixels (i.e, about 0.7X for each dimension), the jaggies go away. (even if I then do a simple rotation with no upsampling etc). Not sure whether it's significant or not, but that corresponds to the total number of green pixels in the sensor.

Also, I downloaded a sample photo from the Nikon D1 which has lots of straight lines, and then did a simple rotation. No jaggies at all.

Greg.

-- Greg Sullivan (gregory.sullivan@compaq.com), January 30, 2000.


I've looked at a lot of sample outdoor portrait photos from higher resolution cameras now, and even they produce jaggies. I guess it makes sense - when looking at photos at 1X magnification, the amount of jaggies should really only depend (mainly) on the CCD pixel layout and the quality of the interpolation. So far I've noticed jaggies on the Nikon Coolpix 950 and the Olympus C2000Z only.

The cameras that do not produce jaggies seem to, nevertheless, blur the edges, mainly in the borders between those vividly coloured flowers. Looking at the Kodak Picture CD photo, the difference is staggering. I always *assumed* that it would be normal for the edges to be a bit jagged or blurred, simply due to monitor resolution. After seeing how good the Picture CD photo is, however, it seems that this imperfection is occuring due to the digital cameras.

It would be good if yet another test photo could be included in the reviews, to really test how well cameras can produce smooth borders between vivid colours, which are not purely horizontal or vertical borders. One area of the photo would concentrate on red and blue borders, and another area would concentrate on cyan & magenta & yellow borders. (to cover the two main types of CCDs which are currently in use - RGB and CMY, although I realise that RGB is probably far more common)

Greg.

-- Greg Sullivan (gregory.sullivan@compaq.com), January 31, 2000.



Moderation questions? read the FAQ